Ab initio study of MgH2: Destabilizing effects of selective substitutionsby transition metals
Adel F. Al Alam a, b, *, Samir F. Matar c, d, Naïm Ouaini a
a Holy Spirit University of Kaslik, USEK, P.O. Box 446, Jounieh, Lebanonb University of Balamand, Department of Physics, P.O. Box 100, Tripoli, Lebanonc CNRS, ICMCB, UPR 9048, F-33600 Pessac, Franced Universit�e de Bordeaux, ICMCB, UPR 9048, F-33600 Pessac, France
a r t i c l e i n f o
Article history:Received 2 March 2014Received in revised form15 July 2014Accepted 21 July 2014Available online 1 August 2014
The strong ionicity of H within rutile MgH2 is reduced by selective substitution of Mg by T (¼Fe, Co, Ni,Pd, Pt) using trirutile super-structure host TMg2H6. These novel model systems, as computed in thequantum mechanical framework of density functional theory, showed a gradual decrease of the chargescarried by H down to �0.02e improving the use of MgH2 for applications.
Hydrogen storage materials such as hydrides are leading can-didates for clean energy in the future. Archetype hydride MgH2 hasbeen studied intensively owing to its large gravimetric density~7.6 wt.%. However, its high thermodynamic stability preventshydrogen absorption/desorption at mild conditions, whence thedifficulty of its ad hoc use in applications. Experimental and theo-retical efforts have culminated over decades to overcome the latterintricacy. The kinetics of hydrogenation were improved experi-mentally either by the addition of catalysts [1e3] or by the intro-duction of nickel as an adjoined metal such as in Mg2NiH4 [4].Recent theoretical investigations suggested the insertion of lightelements such as carbon and boron which decreased the largelyionic character of hydrogen in (B,C)0.167MgH2 [5].
The aim of the present study is to remedy the situation pro-hibiting the use of MgH2 in devices by selective substitution of Mgwith transition metals (T ¼ Fe, Co, Ni, Pd, Pt) in a trirutile hostsuper-structure using first-principles density functional theory(DFT) calculations [6,7].
As illustrated in Fig. 1, ordered trirutile TMg2H6 crystallizes asrutile with the tetragonal structure in space group P42/mnm (No.136). Given inWyckoff letter, T atoms occupy 2a sites at coordinates(0, 0, 0), andMg atoms are found in 4e sites at (0, 0, z~1/3). There aretwo hydrogen sub-lattices, namely H1 at 4f (x, x, 0) andH2 at 8j (x, x,z). Both Mg and T species are surrounded by irregular H octahedra.Successive TeH planes (at z ¼ 0 and z ¼ 1/2) are separated by twoMgeH planes (at z~1/6 and z~1/3).
Archetype MgH2 crystallizes with the tetragonal rutile structurein space group P42/mnm (No. 136). The latter order can becompared to the trirutile structure by substituting T species by Mgat 2a sites, whereby H atoms are located exclusively at 4f sites.
3. Computational methodology
Geometry optimization and total energy calculations wereperformed with the Vienna ab initio simulation package (VASP)[8,9]. The ioneelectron interactions were described using theprojector augmented wave (PAW) method [9,10]. Electronexchange-correlation functionals were built within the generalizedgradient approximation (GGA) scheme following the nonlocalcorrection of Perdew, Burke and Ernzerhof (PBE) [11]. It is impor-tant to mention that semi-core p states where accounted for PAW
Fig. 1. Sketches of the crystal structures of rutile-type MgH2 (left-hand side) and trirutile-type TMg2H6 (right-hand side).
A.F. Al Alam et al. / Solid State Sciences 36 (2014) 47e5148
potentials of Mg in order to obtain the correct physical bulk prop-erties and electronic structures for TMg2H6 models. The conjugate-gradient algorithm [12] is used in this computational scheme torelax the atoms and to optimize the structural parameters until theforces on all the unconstrained atoms were less than 0.02 eV/Å andall stress components were less than 0.003 eV/Å3. The tetrahedronmethod with Bl€ochl corrections [10] and a MethfesselePaxton [13]Gaussian smearing scheme were applied for both geometry relax-ation and to accelerate the total energy calculations. Brillouin-zone(BZ) integrals were approximated using the special k-point sam-pling. The calculations are converged at an energy cut-off of 404 eVfor the plane-wave basis set with respect to the k-point integrationwith a starting mesh of 4 � 4 � 4 up to 8 � 8 � 8 for bestconvergence and relaxation to zero strains.
In this work, the atomic charge of hydrogen is calculated using aBader charge analysis [14]. The latter approach partitions thecontinuous electron density into region bounded by the minima ofthe charge density. Such an analysis can be useful when trendsbetween similar compounds are examined; it does not constitute atool for evaluating absolute ionizations. Bader's analysis is doneusing a fast algorithm operating on a charge density grid [15]. Theresults of computed charges Q are such that they lead to neutralitywhen the respective multiplicities are accounted for.
4. Results and discussions
4.1. Geometry optimization, cohesive energies and hydrogen chargedensity
In as far as TMg2H6 models are novel theoretical models chosenherein, geometry optimization was firstly performed. Starting andoptimized structural parameters are given in Table 1. Rutile-typeMgH2 was also examined to establish trends of stability for thecomputed TMg2H6models. The calculated structural parameters for
MgH2 arewithin2%of the experiment. All TMg2H6models relaxed inthe trirutile structure. The stability of thesemodels can be examinedfrom the computed total electronic energies given in Table 2.
The cohesive energies of various TMg2H6 structures werecalculated with the expression.
Ecoh: ¼ EðT2Mg4H12Þ � 2EðTÞ � 4EðMgÞ � 6EðH2ÞThe energy terms on the right-hand side of the equation
represent, in order, trirutile hydride model, pure T metal, pure Mg,and gas-phase hydrogen. The strength of cohesive energy of amodel is a measure of the stability of that model. Largely negativeEcoh. indicate stable binding, whereas positive energies correspondto an unstable model. The energy of the gas-phase hydrogen dimerwas calculated with an 8� 8� 8 cell. The cohesive energy per H2 ofMgH2 is calculated within 8% of the experimental value �0.79 eV[16]. The computed Ecoh. per H2 values in Table 2 clearly indicatethat all TMg2H6 models are stable owing to the negative values.Compared to MgH2, all models are less stable. The latter findingmeets with the aims of this study in as far as less thermodynami-cally stable hydrides are sought.
The latter should be comforted further by examining the atomiccharge of hydrogen shown as a function of T species in Fig. 2. Asexpected, hydrogen exhibits a less ionic character near T elements(H1 sub-lattice) compared to H charges near Mg (H2 sub-lattice).This can be explained by the electronegativity value of thedifferent species given in the Pauling scale: c(Mg) ¼ 1.31,c(Fe) ¼ 1.83, c(Co) ¼ 1.88, c(Ni) ¼ 1.91, c(Pd) ¼ 2.2, andc(Pt) ¼ 2.28. All T elements are more electronegative than Mg,whence the less ionic hydrogen in their surroundings. Furthermore,H1 charge near T elements undergoes gradual reduction of its ioniccharacter from H�0.4 for FeMg2H6 model down to H�0.02 forPtMg2H6. The latter value is also due to the large H1-Pt separationdH1-Pt ¼ 1.80 Å. The other hydrogen sub-lattice, namely H2, exhibitsa constant evolution around a charge of -6e. Nevertheless the
Table 1Optimized and (starting experimental when available) structural parameters for MgH2 and TMg2H6 models. The distance separating H1 sublattice from T element is also given.
A.F. Al Alam et al. / Solid State Sciences 36 (2014) 47e51 49
overall changes brought by T are established. This substantialreduction of the carried charge by H predicted theoretically shouldbe an indication of the readiness of H desorption experimentally. Itis important to mention that smaller amounts of metal would beneeded to strongly modify MgH2 especially with platinumwhich isan expensive metal. Tests are underway with experimental groupsat our Institute.
Further we comment on the relative changes of charges on Tusing the Bader charge analysis. The values are as follows:
Fe þ 0.71; Co: þ0.50: Ni:þ0.52: Pd: 0.27; Pt: þ0.13. The ioni-zation degree follows closely the electronegativity magnitude withthe following trend: the least electronegative is the least charged.
4.2. Electronic density of states: DOS analysis
Fig. 3 shows the site projected density of states (PDOS) corre-sponding to Mg3H6 and TMg2H6. The energy along the abscissa axisis brought to EV, top of the valence band (VB) which is separatedfrom the conduction band (CB) by a band gap of ~3.5 eV for trirutile-Mg3H6 (Fig. 3a and b). This agrees with the insulating character ofarchetype MgH2 showing a band gap of ~5.6 eV [17]. The VB isdominated by H (H1 and H2) with prevailing H2 intensities due totheir higher multiplicity with respect to H1. Magnesium PDOS aredominating within the CB due to their low filling and electrondeparture towards H.
The DOS's of TMg2H6 show a few similar feature with MgH2-likeDOS within the 8 eV range from �2 to �10 eV (Fig. 3c,d,e). Similar
Table 2Total electronic energies and cohesive energies in units of eV for all TMg2H6 models.
DOS skylines are also observed within the CB between itinerant Tstates and the Mg/H states. The band gap has decreased down to~2.5 eV for Fe due to the covalent character brought in by iron. FromCo to Ni and Pt the extra electrons brought by the increasing Znumber shifts EF to the states Mg and H formerly found within theCB. However the doping is far too high and leads to closing of thegap. A peculiar feature appears for the localized (sharp) T nd-statesPDOS which signals little mixing with the host Mg and H states andcould be labeled as non bonding, as shown by the small PDOSmagnitude of Mg and H below the Fe(d) PDOS for instance. Theinsulating character is preserved is as far as the energy is stillreferred the top of the VB but the gap is much reduced down to~0.3 eV. Clearly the amount of transition metal is too large(FeMg2H6^Fe0.333Mg0.6667H2) by experimental standards. Note
Fig. 2. Atomic charge of hydrogen QH as function of T species in all TMg2H6 models forH1 and H2 sub-lattices. All values of QH are given as a multiple of elementary charge(e ¼ 1.6 � 10�19 C).
Fig. 3. Site projected electronic density of states of: (a) rutile-MgH2, (b) trirutile-Mg3H6, (c) FeMg2H6, (d) CoMg2H6, (e) NiMg2H6, (f) PdMg2H6, (g) PtMg2H6.
A.F. Al Alam et al. / Solid State Sciences 36 (2014) 47e5150
that some of the other transition elements candidates have showntotal metallization, i.e. a closing of the band gap. Then smalleramounts of T elements should be introduced in order to preservethe insulating properties of MgH2 and future works are planned.Nevertheless, our approach using trirutile host structure has shownrelevant effects brought by T substitution on the electronic struc-ture of MgH2 owing to the use of trirutile super-structure allowingselective substitutions of Mg.
5. Conclusion
The use of trirutile host super-structure allowed selective sub-stitution of Mg by T elements. This brought significant effectsrelevant to the reduction of the strong ionic character of H whichprohibited the use of rutile MgH2 in applications. The introductionof T species tends to narrow down the band gap of MgH2 leading to
total metallization. Then smaller amounts should be introduced inorder to preserve the insulating properties of MgH2 as future worksare planned.
Acknowledgment
We acknowledge financial support from French-LebaneseCEDRE project and CSR-USEK. Part of the calculations where doneon MCIA super computers of the University Bordeaux 1. Supportfrom the Conseil R�egional d'Aquitaine is gratefully acknowledged.
A.F. Al Alam et al. / Solid State Sciences 36 (2014) 47e51 51
[3] M. Nakhl, M. Zakhour, Ch. Amine, H. El-Rassy, S.F. Matar, Adv. Mater. Res. 324(2011) 119.
[4] J. Huot, J.F. Pelletier, G. Liang, M. Sutton, R. Schulz, J. Alloys Comp. 330e322(2002) 727.
[5] S.F. Matar, Comp. Mater. Sci. 69 (2013) 424.[6] P. Hohenberg, W. Kohn, Phys. Rev. B 136 (1964) 864.[7] W. Kohn, L.J. Sham, Phys. Rev. A 140 (1965) 1133.[8] G. Kresse, J. Furthmüller, Phys. Rev. B 54 (1996) 11169.[9] G. Kresse, J. Joubert, Phys. Rev. B 59 (1999) 1758.
[10] P.E. Bl€ochl, Phys. Rev. B 50 (1994) 17953.[11] J. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865.
[12] W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetterling, Numerical Recipes,Cambridge University Press, New York, 1986.
[13] M. Methfessel, A.T. Paxton, Phys. Rev. B 40 (1989) 3616.[14] R. Bader, Chem. Rev. 91 (1991) 893.[15] W. Tang, E. Sanville, G. Henkelman, J. Phys.: Condens. Matter 21 (2009)
084204. http://theory.cm.utexas.edu/henkelman/research/bader.[16] M. Yamaguchi, E. Akiba, Materials Science and Technology, vol. 3B, VHC, New
York, 1994, p. 333.[17] C. Moys�es Araújo, S. Leb�egue, O. Eriksson, B. Arnaud, M. Alouani, R. Ahuja,
